From: ara.t.howard@...
Date: 2006-04-14T00:31:03+09:00
Subject: Re: Range#size

On Fri, 14 Apr 2006, Jon Bauman wrote:

> Jeremy Tregunna wrote:
>> On 13-Apr-06, at 9:20 AM, Meinrad Recheis wrote:
>>
>>> does anyone object to adding the method :size to class Range?
>>
>> Forgive me for being blunt, but Ranges don't have a fixed size, so
>> how could you possibly measure it accurately? Force class
>> implementers to implement not only a #succ method for their objects,
>> but also a #range_size method which would give a formula for
>> calculating the size? I mean seriously...
>
> You know, it's not as though all range operations work with all range
> types anyway:
>
> $ irb --simple-prompt
>>> (1.0..2.0).each {|x| p x }
> TypeError: can't iterate from Float
>        from (irb):1:in `each'
>        from (irb):1
>
> So, what would be wrong with allowing size for object that implement the
> - operation and giving an error for those that don't?

nothing.  but try to write it.  it's harder than you might think.

you need to define as something like this:

   class Range
     def size
       begin
         first.distance last
       rescue NoMethodError
         begin
           last.distance first
         rescue NoMethodError
           raise NoMethodError, "size"
         end
       end
     end
   end

   class Fixnum
     def distance other
       (other - self).abs
     end
   end

   class Float
     def distance other
       Float::MAX/Float::MIN # Infinity
     end
   end

in otherwords the endpoints must to the range size checking - otherwise the
Range class would need to know about every class and how to calculate sizes
for all M x N combinations.

if the endpoints calculate it we can easily do

   (0 .. 1).size

or

   (0.0 .. 1.0).size

but what about

   (0 .. 1.0).size

??

how does Range now which endpoint to use as 'other'?  if we use 1.0 it works.
if we use 0 it does not.

it's easy enough to fix this for only Fixnum and Float using a smart coerce -
but then we have to deal with Rational, Bignum, Complex, String, etc.  and we
haven't even hit user defined types.

i'm not saying there isn't an elegant solution.  but the problem becomes
evident imediately when one trys to solve the problem generically.

regards.

-a
-- 
be kind whenever possible... it is always possible.
- h.h. the 14th dali lama